CBO MEMORANDUM

CONGRESSIONAL BUDGET OFFICE

CBOMEMORANDUM

THE EFFECTS OF MANAGED CAREAND MANAGED COMPETITION

February 1995

CBOMEMORANDUM

THE EFFECTS OF MANAGED CAREAND MANAGED COMPETITION

February 1995

CONGRESSIONAL BUDGET OFFICESECOND AND D STREETS, S.W.

WASHINGTON, D.C. 20515

This Congressional Budget Office (CBO) memorandum updates and revises anearlier CBO report—a March 1994 memorandum titled "Effects of Managed Care:An Update." It presents CBO's current assessment of the effects of healthmaintenance organizations based on an analysis of the 1992 National HealthInterview Survey. In keeping with CBOfs mandate to provide impartial analysis, thereport makes no recommendations.

Sandra Christensen of CBO's Health and Human Resources Division wrotethe memorandum under the direction of Nancy Gordon and Linda Bilheimer. RogerFeldman, Marsha Gold, Scott Harrison, Jeffrey Lemieux, Murray Ross, John Sheils,David Stapleton, and Christina Witsberger offered helpful suggestions.

Sherwood Kohn edited the manuscript and Christian Spoor provided editorialassistance. Sharon Corbin-Jallow prepared the final version of the manuscript.Questions about the analysis may be addressed to the author at (202) 226-2665.

CONTENTS

SUMMARY v

INTRODUCTION 1

EFFECTS OF HMOs AND OTHER MANAGEDCARE ARRANGEMENTS 1

Average Effects of HMOs on Use of Services 2Any-Willing-Provider and Point-of-Service Requirements 6Revised Estimates of Effects on Total Spending 8Quality of Care and Enrollee Satisfaction 12

EFFECTS OF MANAGED COMPETITION 14

APPENDIX: THE EFFECTS OF MANAGED CARE ON USEOF SERVICES: EVIDENCE FROM NATIONALSURVEY DATA 18

TABLES

1. Average Reduction in Use Estimated for People in HMOsCompared with Those in Typical Indemnity Plans 3

2. Average Reduction in Use Estimated for People in ManagedCare Plans Compared with Those in Unmanaged IndemnityPlans 9

3. Estimated Savings as a Percentage of Potentially ManageableExpenditures, 1990 10

4. Estimated Savings as a Percentage of Alternative HealthExpenditure Totals, 1990 11

A-l. Variable Definitions and Weighted Means 21

A- 2. Regression Estimates for Outpatient Visits 22

A-3. Regression Estimates for Inpatient Days 23

A-4. Estimated Change in Use of Services for HMOsCompared with Typical Indemnity Plans 24

A-5. Previous Estimates of Change in Use of Servicesfor HMOs Compared with Typical Indemnity Plans 28

A-6. Comparison Under Alternative Specificationsof Estimated Change in Use of Services forHMOs Compared with Typical Indemnity Plans 29

IV

SUMMARY

New evidence from the 1992 National Health Interview Survey indicates that healthmaintenance organizations (HMOs) reduce use of health care services by an averageof about 8 percent, compared with services that similar patients would be expectedto use in a typical fee-for-service indemnity plan. Most of this effect is generatedby group- or staff-model HMOs, which reduce use of services by nearly 20 percent.Although independent practice associations (IPAs) having certain characteristicscould be as effective as group/staff HMOs, many IPAs do not have thesecharacteristics. Further, IPAs would be unable to develop two of the necessarycharacteristics in states that have "any-willing-provider" laws, which require networkplans to accept all providers who are willing to meet the plans1 terms.

The effects on use of services presented in this memorandum, however, donot necessarily imply similar effects on health care spending if enrollment increasedin more tightly managed plans. Spending changes would mirror changes in use onlyif it was reasonable to assume that there would be no alteration in payment rates forproviders or in administrative costs, at least on average. Those assumptions wouldnot be reasonable in many instances. For example, if enrollment in HMOs or othermanaged plans increased, average payment rates for providers might fall becausemanaged plans are more likely than unmanaged plans to negotiate for and win pricediscounts from providers. It is also likely that administrative costs would increase.

INTRODUCTION

The Congressional Budget Office (CBO) last summarized its assessment of researchon the effects of various forms of managed care in a memorandum released in March1994.] That memorandum also presented estimates of national health spending underalternative assumptions about the proportion of insured people who were moved intomore effective forms of managed care.

This memorandum presents CBOfs current assessment of the effects of healthmaintenance organizations (HMOs) based on an analysis of the 1992 National HealthInterview Survey conducted by the National Center for Health Statistics. Further, itexpands the previous memorandum to discuss the effects of point-of-service and"any-willing-provider" requirements on HMOs and other network plans. It alsoexamines the experiences of some purchasing cooperatives that have attempted tointroduce more competitive pressures on health plans.

EFFECTS OF HMOs AND OTHER MANAGED CARE ARRANGEMENTS

The health insurance market has been changing rapidly in recent years. New formsof managed care (such as independent practice associations and point-of-serviceplans) have developed to compete with group/staff HMOs, and even most indemnityplans have now incorporated some elements of managed care.2 Participation in allforms of managed care has increased as both providers and patients have becomemore accepting of limits imposed by insurers.

How much the growth of managed care has reduced the use medical servicesand health care costs is uncertain. Data and methodological limitations make itdifficult to isolate the effects of specific types of managed care. Nevertheless,estimates of these effects are important for assessing some legislative proposals. Inthe comparisons made in this memorandum, a "traditional" indemnity plan is one thathas no managed care component. A "typical" indemnity plan is one that has someelements of managed care—such as utilization review or a network of preferredproviders.

1. Congressional Budget Office, "Effects of Managed Care: An Update," CBO Memorandum (March1994).

2. An independent practice association is an HMO that contracts with individual physicians or groupsto provide services to its enrollees in the physicians' private offices. The physicians also continue totreat patients not enrolled in the HMO. Physicians in group/staff HMOs treat the HMO's patientsexclusively. Until recently, independent practice associations and group/staff HMOs were generally"closed-panel" plans, in which enrollees were restricted to the plan's panel of providers. In recentyears, however, the "open-panel" or point-of-service option has become more common. Under thatoption, plans pay part of the costs of covered services from out-of-plan providers.

Average Effects of HMOs on Use of Services

In CBOfs 1994 managed care memorandum, assumptions about the potential drop inuse of health care services under proposals that would shift insurance enrollment toHMOs were based on a study by Lewin-VHI, Inc., which used data for 1989 from theNational Health Interview Survey (NfflS).3 New evidence from the 1992 NfflS,however, indicates that the Lewin-VHI results probably understate the reduction inuse to be expected from HMOs when compared with indemnity plans. Byintroducing a variable for childbirth into the equations, CBO was able to control foran important source of adverse selection among HMO enrollees that apparentlybiased the results from the Lewin-VHI study.

CBOfs findings from the 1992 NfflS indicate that the reduction in total useof services for HMOs is about twice as large as the estimates reported by Lewin-VHI.(See the appendix for CBO's analysis.) Compared with the current mix of managedand unmanaged indemnity plans, group/staff HMOs reduce use of medical servicesby an estimated average of 19.6 percent, instead of the 9.1 percent reduction impliedby the Lewin-VHI results; independent practice associations (IPAs) reduce use byan average of about 0.8 percent, instead of 0.3 percent (see Table 1). When thesefindings for group/staff HMOs and IPAs are combined using 1992 year-endenrollment patterns, the estimated average effect of HMOs is to reduce use ofservices by 7.8 percent, when compared with the current mix of indemnity plans.4

The 1994 memorandum assumed that the average effect was about 4 percent.5

The results in this memorandum—which focus on use of services, notcosts—confirm the Lewin-Vffl findings that reductions in use of services from HMOscurrently come almost entirely from group/staff models. On average, IPAs do littlebetter than indemnity plans. As discussed in CBO's March 1994 memorandum, theIP A form of HMO can be as effective as group/staff HMOs if certain conditions aremet, but frequently they are not. The IPAs that are most likely to approach theeffectiveness of the best group/staff HMOs are selective about using cost-consciousproviders, maintain an effective network for information and control, place providersat financial risk, and generate a substantial portion of each provider's patient load.

3. Lewin-VHI, Inc., The Financial Impact of the Health Security Act, Appendix A (Fairfax, Va: Lewin-VHI, Inc., December 9, 1993), Table A-4; and Lewin-Vffl, Inc., "Effects of Managed Care,Uninsurance, and AIDS on Health Care Use" (Fairfax, Va.: Lewin-Vffl, Inc., February 15,1993).

4. According to a 1993 report by the Group Health Association of America (Patterns in HMOEnrollment, p. 25), 37 percent of HMO enrollees were in group/staff models at the end of 1992, and63 percent were in IPAs and network plans.

5. In addition to the different effects estimated for group/staff HMOs and IPAs, this average used the1989 distribution of HMO enrollment-41 percent in group/staff models and 59 percent in IPAs.

TABLE 1. AVERAGE REDUCTION IN USE ESTIMATED FOR PEOPLE IN HMOsCOMPARED WITH THOSE IN TYPICAL INDEMNITY PLANS

Percentage Reduction in Useof Services bv Tvoe of HMO

Current Estimates

Previous Estimates

Group/Staff

19.6

9.1

IPA

0.8

0.3

HMO Average

7.8a

3.9b

. SOURCE: Congressional Budget Office.

NOTES: Typical indemnity plan refers to a fee-for-service plan with some elements of managed care.

HMO = health maintenance organization; IPA = independent practice association.

a. Assumes 37 percent in group/staff HMOs and 63 percent in IPAs.

b. Assumes 41 percent in group/staff HMOs and 59 percent in IPAs.

A study of Aetna, Humana, and Prudential insurance plans confirms theimportance of patient volume if IP As are to affect providers1 treatment patterns.Managers in these plans said they could not get the attention of network physiciansunless the physician had at least 100 patients from the plan. Regression estimatesfrom Humana plans indicated that an increase in volume from 100 enrollees to 1,000per physician was associated with a 6.4 percent decline in specialty and hospital carecosts per enrollee. The effect on costs was especially large up to about 600 patientsper physician.6

It is noteworthy that the HMO effects discussed above refer to differences inuse of services between the typical HMO and the typical indemnity plan in 1992.Because indemnity plans usually have higher cost-sharing requirements than doHMOs, and because cost sharing reduces patients1 use of services, the isolated effectof the HMO style of practice on reducing use of services is greater than the estimatedeffects show. In other words, if HMOs and indemnity plans imposed the same cost-sharing requirements, the difference in use of services between enrollees in HMOsand those in indemnity plans would be larger than the estimates given above. Formost proposals, however, there is no reason to believe that the dissimilar cost-sharingrequirements for HMOs and indemnity plans would change, so the smaller effectsestimated from the NHIS for typical plans are the relevant ones.

The assessment in this memorandum about the savings to be expected if moreinsured people were in managed care plans refers to effects on use of services only,compared with the expected use of services by similar patients in a typical indemnityplan. Additional effects (either savings or costs) may be caused by changes inprovider prices or discounts and in administrative costs, but this analysis makes noattempt to incorporate them. CBO's cost estimates for specific health legislation,however, do incorporate estimates of these effects.

Confusion sometimes arises when analysts report effects based on differentconcepts. This can be illustrated by referring to a recent study released by Lewin-VHI.7 The study reports that costs for patients in Aetna's IP As were 23 percentlower than costs for similar patients in traditional indemnity plans, after eliminatingthe effects of benefit and copayment differences. To make this estimate comparableto CBO's estimated reduction in use of 0.8 percent for IP As compared with thetypical indemnity plan, however, three adjustments must be made. In particular, itis necessary to:

6. See David C. Stapleton, "New Evidence on Savings from Network Models of Managed Care" (Fairfax,Va.: Lewin-Vffl, Inc., May 1994).

7. Ibid.

o Eliminate the savings attributable to the IPA's negotiated price discounts bysubtracting 15 percent;8

o Subtract 4 percent to reflect the savings that indemnity plans typicallyachieve through utilization review;9 and

o Subtract an additional 4 percent to account for the relatively low cost-sharingrequirements in IP As.10

The first adjustment eliminates savings from price discounts because CBO'sestimate from the NHIS considers only effects on use of services. (Savings or coststhat might arise from price discounts or any changes in price are, however,incorporated into CBO's estimates for specific legislative proposals in a separatestep.) The second adjustment is necessary because Aetna's estimate compares IP Aswith traditional (unmanaged) indemnity plans, and CBOfs estimate compares IP Aswith typical indemnity plans (which now generally include elements of managedcare, such as utilization review). The third is necessary because IPAs typically havelower cost-sharing requirements than do indemnity plans, and Aetna's estimate wasdesigned to eliminate the effects of these differences. If cost-sharing requirementswere identical for the two kinds of plans, Aetna's results would indicate that IPAsreduce use of services by about 4 percent compared with its indemnity plans.11

However, the lower cost-sharing requirements that are typical of IPAs encouragegreater use of services by enrollees.

Applying the three adjustments described above to Aetna's original estimateof savings of 23 percent yields an estimate comparable to CBO's, and it indicatesthat the typical DP A does not appreciably reduce use of services when compared with

8. Aetna estimated the average provider discount in its network plans at 15 percent.

9. Aetna estimated savings of 4 percent from utilization review in its indemnity plans.

10. The amount needed to reflect the lower cost-sharing requirements in Aetna's typical IP A, comparedwith its indemnity plans, was estimate^ based on the "low" and "high" cost-sharing amounts shownin Exhibit II. 1 of the Stapleton paper. The copayment for an office visit under the "low" IPA plan wastwo-thirds of the copayment under the "high" plan ($10 versus $15); consequently, it was assumedthat average cost-sharing requirements in Aetna's IPAs (low cost-sharing plans) were about two-thirdsof average cost-sharing requirements in Aetna's indemnity (high cost-sharing) plans. To eliminateAetna's adjustment to equalize cost sharing, it was estimated that use of services for IPA enrolleeswould be about 4 percent higher if cost sharing was reduced from an average of 25 percent for highcost-sharing plans to about 17 percent, assuming an elasticity of -0.1. This adjustment could besmaller depending on the proportion of indemnity enrollees whose plans offer them lower cost sharingwhen they use the plan's preferred panel of providers.

11. Results are derived using Aetna's 23 percent estimate minus the 15 percent adjustment to eliminate theeffects of fee discounts and minus die 4 percent adjustment to recognize the utilization review nowtypical for indemnity plans.

the typical indemnity plan. By utilization review, both kinds of plans can reduce useof services by about 4 percent when compared with a traditional (unmanaged)indemnity plan. Independent practice associations might reduce use by another 4percent by applying other provider controls, but in the typical IPA that effect isgenerally offset by the higher patient use encouraged by the IPA's low cost-sharingrequirements. Independent practice associations also typically reduce enrollees'health care costs by negotiating discounted prices from providers, but that is beyondthe scope of this analysis, which focuses entirely on effects on use of health careservices.

Any- Willing-Provider and Point-of-Service Requirements

In some states, HMOs and network health plans are subject to "any-willing-provider11

(AWP) requirements, which prohibit network plans (IP As and preferred providerorganizations, but not group/staff HMOs) from excluding any providers in the areawho are willing to accept a plan's terms for participation in the network. A variantof this is the point-of-service (POS) requirement, under which health plans havinga limited group of providers (group/staff HMOs and IPAs) are required to reimburseenrollees for covered services received from out-of-plan providers. Generally,however, plans are permitted to impose higher cost-sharing requirements on out-of-plan services, similar to the different cost-sharing requirements that preferredprovider organizations (PPOs) impose in order to encourage enrollees to use theplans' preferred providers.

AWP requirements are intended to ensure that no providers are denied accessto a plan's patient population if they are willing to abide by the plan's conditions forparticipation. POS requirements are intended to ensure that patients are free to usethe providers of their choice (although they may have to pay more of the costs).Proponents of managed care believe, however, that these requirements greatly reducea managed care plan's ability to provide lower-cost care than traditional indemnityplans. Proponents of the restrictive requirements argue that the protection theyprovide patients and providers is of sufficient importance to justify the decreasedability of managed plans to reduce costs. Thus, the policy decision to impose AWPor POS requirements on health plans comes down to a value judgement: How muchis society willing to pay in higher health care costs to protect patients and providersin this way?

Any- Willing-Provider flgqfflj^mgntS- In the states where they exist, theserequirements apply to plans that have nonexclusive contracts with a network ofproviders (IPAs and PPOs), but not to group/staff HMOs that employ physiciansexclusively to serve the plan's patients. Thus, the most effective type of HMO is not

subject to AWP requirements. The potential for IP As to constrain use of services,however, may be adversely affected by AWP requirements.

Although the evidence to date indicates that the typical IPA does notsignificantly alter use of services compared with indemnity plans, IP As with certaincharacteristics have the potential to reduce use as effectively as group/staff HMOs.An AWP requirement, however, might make it impossible for IP As to develop twoof those necessary characteristics: selective contracting with cost-effective providersand sufficient IPA patient volume to affect providers' behavior. Further, having tocontract with all interested physicians in a region increases an IPA's costs in twoother ways: the IPA incurs higher administrative costs because it has to contract witha larger number of providers and document disenrollment decisions more thoroughly,and it is less able to negotiate price discounts with providers because it cannot assureeach participating provider enough of an increase in volume to offset the discount.

The Point-of-Service Option. In 1992, about 50 percent of HMO plans provided aPOS option for enrollees, although these plans covered only about 6 percent of allHMO enrollees. Only 15 percent of those enrolled in a POS option were ingroup/staff HMOs; the rest were in IPAs or other network-model HMOs.12

There is little solid information about the effects that a POS option wouldhave on a plan's ability to control use of services, although they could be substantialeven if only a small proportion of enrollees used out-of-plan providers. The reasonis that those patients who are most likely to consult physicians outside the plan'snetwork have serious health problems and want treatment from particular specialists.Because the highest-cost 5 percent of enrollees typically account for 50 percent ormore of health plan costs, a few of these patients can transfer a substantial portion ofthe plan's benefit costs to out-of-plan providers.

Plans could, however, counteract the costly effects of a POS option byimposing sufficiently high cost-sharing requirements on out-of-plan use. Indeed,evidence from Aetna indicates that its IPA-POS plans were able to reduce benefitcosts by about as much as its closed-panel IPAs, when compared with a traditionalindemnity plan, but the source of the reduction in benefit costs may be verydifferent.13 As explained earlier, most of Aetna's savings from IPAs result fromdiscounts on providers' fees, and there is little effect on use of services beyond thatachieved by utilization review. POS plans will gain no savings from discounts on

12. Group Health Association of America, Patterns in HMO Enrollment (Washington, D.C.: Group HealthAssociation of America, 1993).

13. Stapleton, "New Evidence on Savings," Exhibit H.3.

7

out-of-plan services, so the loss of discounts must be made up from greater patientcost sharing and less use of services. Moreover, the IPA-POS plans have higheradministrative costs than do closed-panel IP As.

Revised Estimates of Effects on Total Spending

Revised estimates of the effects of HMOs imply that those to be expected for eachperson moving from an indemnity plan to an HMO would be greater than waspredicted in CBO's 1994 memorandum. (Tables 2 through 4 revise the resultsfor 1990 that were shown in Tables 4 through 6 of that earlier memorandum.)Compared with a traditional (unmanaged) indemnity plan, CBOfs most recentfindings indicate that group/staff HMOs reduce use of services by 21.9 percent forprivately insured patients, and IP As reduce use by an average of 3.6 percent (seeTable 2).14 Assumptions about the effects of managed indemnity plans on use ofservices are unchanged from the previous memorandum. Compared with unmanagedcare, effective utilization review (including precertification and concurrent reviewfor hospital stays) can reduce use by 4 percent, while other, less effective forms ofmanagement in indemnity plans are assumed to reduce use by about 2 percent.15 In1990, about 47 percent of private indemnity plans had effective utilization reviewprograms in place, 47 percent had less effective forms of use management, and 6percent were unmanaged; consequently, use of services in the indemnity sector wasabout 2.8 percent lower than it would have been if all indemnity plans had beenunmanaged.

Under these assumptions about changes in use of services, and assuming thatspending would mirror changes in use, spending on insured services would have beenlower by 16.6 percent nationwide if all insured people had been enrolled ingroup/staff or other equally effective HMOs in 1990 (see Table 3). Total nationalhealth expenditures would have been lower by 11.9 percent (see Table 4).

Because some people live in areas not populous enough to support HMOs,however, it is unrealistic to assume that all insured people could be enrolled ineffective HMOs, even after the lengthy period needed for their development. If,instead, all insured people in less effective forms of managed care had been moved

14. These results are the combined effects for HMOs compared with the current mix of managed andunmanaged indemnity plans, augmented by the estimated average reduction in use (2.8 percent)generated by the current mix of indemnity plans compared with an unmanaged plan.

15. The 4 percent assumption is consistent with Aetna's estimated effects of utilization review in itsindemnity plans.

8

TABLE 2. AVERAGE REDUCTION IN USE ESTIMATED FOR PEOPLE INMANAGED CARE PLANS COMPARED WITH THOSE INUNMANAGED INDEMNITY PLANS

Percentage Reduction in Use of Services byPrimary Source of Type of Managed Care Arrangement8

Insurance Coverage I II III IV V

MedicareMedicaidPrivate or Other PublicNo Insurance

21.910.921.9

0

3.61.83.6

0

4.02.04.0

0

2.01.02.0

0

0000

SOURCE: Congressional Budget Office.

NOTE: The effect of managed care on use of services for Medicaid enrollees is half the reduction assumedfor enrollees in Medicare and private insurance plans, reflecting the expectation that payment ratesand access to services for Medicaid enrollees would increase under managed care arrangements.

a. Categories of managed care:I. Group/staff model health maintenance organizationsn. Independent practice associationsm. Utilization review including precertification and concurrent review of hospital careIV. Other forms of managed careV. Unmanaged (traditional) fee-for-service indemnity plans

TABLE 3. ESTIMATED SAVINGS AS A PERCENTAGE OFPOTENTIALLY MANAGEABLE EXPENDITURES, 1990

All Sources

Distribution of InsuredPopulation bv Type of Managed Care

Primary Source ofInsurance Coverage

All inEffectiveHMOs

Six Percentin EffectiveHMOs and

94 Percent inEffective UR

Seventy Percentin EffectiveHMOs and


16.6 1.3 11.7

MedicareMedicaidPrivate or Other PublicNo Insurance

SOURCE: Congressional Budget Office.

NOTES: Potentially manageable expenditur

19.510.718.1

0

esarethei

1.91.91.0

0

portion of health care sp

14.08.0

12.50

rending that manajged care couldaffect, which includes all personal health services that are typically offered as insurance benefits.The analysis assumes that changes in use result in comparable changes in spending, althoughpayment methods and differences in administrative costs may preclude this.

HMO = health maintenance organization; UR = utilization review.

10

TABLE 4. ESTIMATED SAVINGS AS A PERCENTAGE OFALTERNATIVE HEALTH EXPENDITURE TOTALS, 1990

Manageable PersonalHealth Care Expenditures

All PersonalHealth Care Expenditures

National Health Expenditures

Distribution of InsuredPopulation bv Tvoe of Managed Care

Form of HealthExpenditure

All inEffectiveHMOs

Six Percentin EffectiveHMOs and


Seventy Percenthi EffectiveHMOs and


16.6

13.5

11.9

1.3

1.0

0.9

11.7

9.5

8.3

SOURCE: Congressional Budget Office

NOTES: Potentially manageable expenditures are the portion of health care spending that managed care couldaffect, which includes all personal health services that are typically offered as insurance benefits.The analysis assumes that changes in use result in comparable changes in spending, althoughpayment methods and differences in administrative costs may preclude this.

HMO = health maintenance organization; UR = utilization review.

11

into effective utilization review programs, and the proportion of insured people in themost effective HMOs was unchanged (at about 6 percent), spending on insuredservices would have been lower by an estimated 1.3 percent in 1990. National healthexpenditures would have been lower by 0.9 percent.16

Alternatively, if the 70 percent of the population who live in areas populousenough to support effective HMOs had been enrolled in them, and the other 30percent were enrolled in indemnity plans with effective utilization review programs,insured spending would have been lower by an estimated 11.7 percent in 1990.National health spending would have been lower by 8.3 percent.

The assumption in Tables 3 and 4 that spending would mirror changes in useof services implies that average provider and administrative costs would beunchanged from their current levels, but there are reasons why that might not occur.For example, as enrollment shifted to managed care, average administrative costswould probably increase because managed plans tend to have higher administrativeexpenses than unmanaged plans. Average payment rates would fall if managedplans were at least as successful as they are now in negotiating lower rates for theirenrollees, and if providers were unable to recover their lost revenue by increasing therates charged to other payers.

Moreover, although there is evidence that Medicare patients enrolled in risk-based HMOs use fewer Medicare services than they would have used in the fee-for-service sector, Medicare's costs generally increase for each enrollee who switchesto an HMO under the current payment system. This happens because Medicare's percapita payment to the HMO-which is based on what enrollees cost in the fee-for-service sector—does not adequately reflect the generally healthier population thatchooses the HMO option.17

Quality of Care and Enrollee Satisfaction

HMOs appear to provide health care that is roughly comparable with that availablethrough indemnity plans for most conditions, with two broad exceptions-onefavorable and one unfavorable. On the favorable side, HMOs tend to provide more

16. This result is unchanged from the result reported in CBO's 1994 memorandum because the only shiftsthat would occur are into indemnity plans with effective utilization review. Assumed savings for sucha shift are unchanged from the previous memorandum.

17. Jerrold Hill and others, The Impact of the Medicare Risk Program on the Use of Services and Coststo Medicare: Final Version" (Princeton, N.J.: Mathematica Policy Research, Inc., December 3,1992).

12

prenatal, preventive, and cancer-screening services than do indemnity plans.18 Thus,cancers tend to be diagnosed at an earlier, more curable stage.19 On the unfavorableside, group/staff HMOs and DP As that restrict access to specialists sometimes do sowith adverse effects for patients, especially those who have conditions for whichtreatment norms are not well defined. For example, primary care physicians inHMOs are less likely to diagnose or treat patients with depressive disordersappropriately, although treatment is comparable to that provided in indemnity plansonce the patient is referred to a mental health specialist.20

One study concluded, "Restrictive practices do not seem to adversely affectmost people in good health or those with only minor health problems, but thesepractices pose special risks for people whose health is poor. In the RAND HealthInsurance Experiment, low-income individuals and families who were in poor healthat the start of the study and who were randomly assigned to a large, well-establishedHMO had, by the end of the experiment, more bed-days per year due to poor healthand more serious symptoms than those assigned to the fee-for-service plan with nopatient cost sharing, and they had a greater risk of dying than those in the fee-for-service plan with cost sharing."21 A more recent study concluded, "little evidenceexists to show that the successes of prepaid care in relatively healthy populations canbe replicated among sicker patients."22

HMO enrollees tend to be more satisfied with their benefits and premiumsthan people in indemnity plans but less satisfied with their access to and theirinteractions with HMO physicians.23 HMO enrollees in both group/staff models andIP As have better financial access to care because cost-sharing requirements are lowerthan those in indemnity plans. Further, coordination of patient care tends to be better

18. R.H. Miller and H.S. Luft, "Managed Care Plan Performance Since 1980: A Literature Analysis,"Journal of the American Medical Association, vol. 271, no. 19 (May 18,1994).

19. G.F. Riley and others, "Stage of Cancer at Diagnosis for Medicare HMO and Fee-for-ServiceEnrollees," American Journal of Public Health, vol. 84, no. 10 (October 1994).

20. K.B. Wells and others, "Detection of Depressive Disorder for Patients Receiving Prepaid or Fee-for-Service Care: Results from the Medical Outcomes Study," Journal of the American MedicalAssociation, vol. 262, no. 23 (December 15,1989).

21. Thomas Rice, E. Richard Brown, and Roberta Wyn, "Holes in the Jackson Hole Approach to HealthCare Reform," Journal of the American Medical Association, vol. 270, no. 11 (September 15,1993).The RAND results were reported Jn I.E. Ware and others, "Comparison of Health Outcomes at aHealth Maintenance Organization with Those of Fee-for-Service Care," Lancet, no. 1 (1986).

22. D.G. Safran, A.R. Tarlov, and W.H. Rogers, "Primary Care Performance in Fee-for-Service andPrepaid Health Care Systems: Results from the Medical Outcomes Study," Journal of the AmericanMedical Association, vol. 271, no. 20 (May 25,1994).

23. Ibid.

13

in group/staff HMOs and in some IP As (those with primary care case-management)than in indemnity plans. Organizational access is lower for group/staff enrollees,however, because patients have greater difficulty reaching a physician with a medicalquestion and obtaining medical care on short notice. In addition, group/staffenrollees are less likely to have the same physician for subsequent care.

EFFECTS OF MANAGED COMPETITION

The higher estimated savings for HMOs obtained from CBOfs analysis of the 1992NHIS may understate their long-term savings potential. Proponents of managedcompetition believe that a number of factors that are still characteristic of the healthcare marketplace have shielded HMOs from the kind of competition that wouldinduce them to achieve their full potential. Perhaps the most important factor hasbeen the tendency of employers (encouraged by the tax laws) to subsidize generouslywhatever health plan choice employees make, with little or no penalty for choosinga high-cost plan when a menu of plans is offered.

Proponents of managed competition believe that if employees were requiredto pay all of the premium difference between their choice and the lowest-cost plandeemed acceptable, they would rapidly switch to lower-cost plans.24 Consequently,plans would have strong incentives to operate efficiently and to price their benefitslower. By contrast, many believe that under current incentives HMOs need notcharge the lowest cost at which they could provide benefits. Because^manyemployers contribute up to a fixed amount based on the average cost of the mostpopular plans they offer (which tend to be higher-cost indemnity plans), HMOs thatcould offer the specified benefits for a lower premium than the employer's fixedcontribution have no incentive to do so, because it would not increase enrollment.25

The State Employee Insurance Program (SHIP) in Minnesota provides thebest example to date of managed competition in practice.26 The SHIP is a purchasingcooperative that offers to state employees a number of health plans that have similar

24. Economists believe that even the employer-paid share of premiums is effectively paid by employees,in the form of reduced wages, and that in the long run employees' choices reflect this reality.Nevertheless, the share that is nominally paid by the employer is important to employee choices evenin the long run because under current law the employer share is paid from pretax income, whereas thewage income that would otherwise be paid the worker would be taxed. Thus, the employer-paid shareis subsidized through the tax system.

25. Enrollees who choose plans costing less than the employer's fixed contribution amount do not receivethe excess contribution; instead, the employer keeps it

26. See John Klein and Robert Cooley, "Managed Competition in Minnesota," Managed Care Quarterly,vol. 1, no. 4 (Autumn 1993).

14

(but not identical) benefits during an annual open-enrollment period. It providesemployees with comparative information not only about participating plans1

premiums, but also about enrollee satisfaction. Community rating applies, and thestate's contribution toward premiums is set by the lowest-cost plan available to eachemployee.27 Because the program is still evolving and took its current form onlyrecently, however, it is too soon to draw more than tentative conclusions.Furthermore, because it covers only state employees, its effects may differsignificantly from those of a comprehensive program.

Before 1986, Minnesota based its contribution toward health insurance costsfor state employees on SEIP's most popular plan, which was traditional indemnitycoverage offered by Blue Cross. The state paid 100 percent of this premium foremployees, and 90 percent for dependents. Premiums for other plans offered by thestate clustered closely around the Blue Cross premium. Effective for the 1986contract year, however, the state changed the basis for its contribution to the lowest-cost plan available in each area. Employees who chose the lowest-cost plan paidnothing for their coverage, but those who chose a more expensive plan paid theamount by which their plan's premium exceeded that of the lowest-cost plan.

After this change in the basis for the state's contribution, significant variationin the premiums charged by participating plans emerged. The result was a substantialshift in enrollment toward the lowest-priced plans. By 1989, a group/staff HMO hadbecome the lowest-priced plan in the areas it served. Its share of enrollment in theMinneapolis metropolitan area grew from 27 percent in 1988 to 51 percent by 1993(growth statewide was from 19 percent to 35 percent). The unmanaged plan that hadbeen the basis for the state's contribution before 1986 was changed in 1990 to a PPOin an attempt to remain competitive; nevertheless, its share of enrollment in theMinneapolis metropolitan area dropped from 42 percent in 1988 to 16 percent by1993 (statewide it dropped from 56 percent to 42 percent). Thus, enrollees werequite responsive to the price they paid out of pocket for insurance. Further, enrolleeswere willing to shift not only among plans of a similar type, but also among those of

27. See Congressional Budget Office, Managed Competition and Its Potential to Reduce Health Spending(May 1993), Chapter 3, for a description of the important features of managed competition. Of thefeatures listed in the CBO study, the Minnesota program lacks only a risk-adjustment mechanism tooffset any effects of selection bias that may occur among the competing health plans.

15

different types-from higher-cost plans that permitted patients to see any provider oftheir choice to HMOs with a closed panel of providers.28

The rate of growth in premiums for state employees also slowed significantlyafter 1989, although the competitive discipline that resulted from employees1

enrollment choices was not the only relevant influence. The state also subjectedeach plan's rate proposal to analysis by an independent actuary and negotiatedreasonable rate changes where appropriate. Further, the reduction in averagepremium growth for SHIP also reflected the resolution of the financial problems ofthe Blue Cross plan (which introduced a PPO and other elements of managed care)and an economywide slowing in the rate of inflation. The net result of changes toSEIP was to slow the growth in premiums from rates that were above the nationalaverage to rates that were below it. Once having adjusted to a lower level of costs,however, the rate of growth in premiums under SEIP might return to its historicaltrend.

Although proponents of managed competition differ about whether theybelieve purchasing cooperatives should have the authority to negotiate with healthplans about premiums, the ability to negotiate with insurers may be a critical factorin the ability to control costs. Especially in its first few years, the negotiation processin Minnesota sometimes led to significant reductions in rates from those firstproposed. Some insurers were not well prepared to defend their rating methods, ormade errors in rate-setting calculations or assumptions that would have been costlyto the state and its employees if they had gone undetected.

A number of other states also have purchasing cooperatives that incorporatesome elements of managed competition.29 These cooperatives typically have theauthority to negotiate premiums with health plans, and their administrators generallybelieve that this authority is critical in controlling costs. Most introduced ratenegotiation only recently, however, subsequent to experiencing rapid increases inhealth plan premiums. All the cooperatives require plans to submit operating datato justify their premium proposals and some, like SEIP, use independent actuaries todevelop target premiums. If a plan's bid is significantly higher than the target

28. The enrollment shifts that took place in the Minneapolis metropolitan area between 1989 and 1993were consistent with the shifts that would be predicted in response to enrollees* out-of-pocket premiumcosts using the estimates developed by Roger Feldman and others, "The Demand for Employment-Based Health Insurance Plans," Journal of Human Resources, vol. 24, no. 1 (Winter 1989). See alsoRoger Feldman and Bryan Dowd, The Effectiveness of Managed Competition in Reducing the Costsof Health Insurance,11 in R.B. Helms, ed., Health Policy Reform: Competition and Controls(Washington, D.C.: AEI Press, 1993), Chapter 7.

29. See General Accounting Office, Access to Health Insurance: Public and Private Employers'Experience with Purchasing Cooperatives, GAO/HEHS-94-142 (May 1994), for a description ofexisting cooperatives.

16

premium, officials of the cooperative discuss the discrepancies with representativesof the plan and ask them to submit a best and final offer, which is often substantiallylower.

Thus, there is evidence that introducing elements of managed competition,including rate negotiation, can slow the rate of growth in health plan premiums. Forits estimates of health reform proposals in the last session of Congress, CBOassumed that a comprehensive plan that had all the important elements of managedcompetition might ultimately reduce the rate of growth in health spending for peoplein the managed competition sector by 1 percentage point. That reduction was aboutone-quarter of the amount by which growth rates exceeded those attributable topopulation growth and economywide inflation.

Experience under existing systems of managed competition, however, is toolimited to assess accurately the magnitude or duration of effects. For one thing, theintroduction of significant elements of managed competition has occurredsimultaneously with a slowdown in the rate of economywide inflation, so some ofthe apparent effects of managed competition may only reflect that slowdown.Further, recent experiences in Minnesota have raised concerns that the pressuresgenerated by managed competition may trigger too much consolidation among healthplans and providers for genuine competition to thrive. Although the Minnesotalegislature recently passed laws intended to establish a statewide delivery systemcomposed largely of competing "integrated service networks," competitors havecombined at a much more rapid pace than expected. By 1993, about 78 percent ofMinnesotans in managed care plans were enrolled in one of three organizations. Thelegislature has temporarily banned further horizontal mergers involving any of thesethree plans.30

30. Office of Technology Assessment, Managed Care and Competitive Health Care Markets: The TwinCities Experience? OTA-BP-H-130 (July 1994). Also see National Health Policy Forum, "Con-solidation in the Health Care Marketplace and Antitrust Policy," Issue Brief No. 660 (Washington,D.C.: National Health Policy Forum, 1995).

17

APPENDIX: THE EFFECTS OF MANAGED CARE ON USE OF SERVICES:EVIDENCE FROM NATIONAL SURVEY DATA

A number of studies indicate that enrollees in health maintenance organizations(HMOs) use fewer services than similar patients in indemnity plans, with little or noadverse effect on health. Some of the studies also appear to show that theindependent practice association (IPA) form of HMO is as effective as the group/staffmodel. The primary distinction between these two forms is that physicians in IP Astreat a mix of HMO and fee-for-service patients, while those in group/staff modelstreat only patients enrolled in their HMO. The group/staff models are generallythought to be more tightly managed, and thus more effective at controlling use andcosts.

Although there is some evidence that well-managed IP As can control use ofservices as well as group/staff HMOs, data from national surveys indicate that, onaverage, use of services by IPA enrollees is only slightly lower than it is forindemnity patients. Average use of services by enrollees in group/staff HMOs,however, is substantially lower than use by either indemnity or IPA patients.

This memorandum uses data from the 1992 National Health Interview Survey(NHIS) to estimate the average effects of group/staff HMOs and IP As compared withindemnity plans in the fee-for-service sector. It updates and revises earlier resultsreported by Lewin-Vffl, based on the 1989 NHIS.1

The estimates show that average use of medical services by patients ingroup/staff HMOs is about 20 percent lower than that of similar patients in indemnityplans, and use by patients in IP As is about 1 percent lower. Weighting the effects forgroup/staff HMOs and IPAs by 1992 year-end HMO enrollment indicates that, whencompared with indemnity plans, the current mix of HMOs reduces use of services byan average of 7.8 percent.2 These HMO effects are about twice as large as thosefound by Lewin-VHI. One reason for the larger effects is that CBO used informationabout whether or not a respondent gave birth in the hospital at any time during theyear, whereas Levin-VHI did not. Thus, the Lewin-VHI study did not control for an

1. Lewin-Vffl, Inc., The Financial Impact of the Health Security Act, Appendix A (Fairfax, Va.: Lewin-Vffl, Inc., December 9, 1993), Table A-4; and Lewin-Vffl., Inc., "Effects of Managed Care,Uninsurance, and AIDS on Health Care Use" (Fairfax, Va.: Lewin-Vffl, Inc., February 15,1993).

2. In 1992, an estimated 63 percent of HMO enrollees were in IPAs (including network plans) and 37percent were in group/staff HMOs. See Group Health Association of America, Patterns in HMOEnrollment (Washington, D.C.: Group Health Association of America, 1993).

important source of adverse selection affecting HMOs-the preference of manypeople planning a pregnancy to insure through an HMO because of the relativelygenerous maternity and well-baby benefits that HMOs typically provide.

Data and Methods

The data used for this analysis were taken from the 1992 National Health InterviewSurvey conducted by the National Center for Health Statistics, which included ahealth insurance supplement that describes the type of insurance plans eachrespondent had.3 The NHIS is an annual survey of about 130,000 individuals fromabout 50,000 households that are representative of the civilian noninstitutionalpopulation of the United States.4

The sample used here included all NHIS respondents who at the time of thesurvey were less than 65 years old, had private insurance coverage, and did not havepublic insurance (Medicare, Medicaid, or other publicly assisted medical coverage).Those who met the criteria but did not know whether their insurance plan was anHMO or not were excluded. The resulting unweighted sample included 47,822people.

CBO estimated two sets of multivariate regression equations—one to explainrespondents* use of outpatient medical visits during the 12 months before the survey,and the other to explain their use of hospital inpatient days. In each case, CBO usedtwo equations to explain respondents' use of services-logistic regression to estimatethe probability that the respondent had any use (one or more outpatient visits, or oneor more inpatient stays) during the year, and ordinary least squares regression topredict the amount of use (number of outpatient visits for those with any visits duringthe year, or number of inpatient days for those with at least one hospital admission).The predicted probability of any use multiplied by the predicted amount of use forusers gives an estimate of the total amount of use for a respondent with a given setof characteristics.

The explanatory variables used here were the same for each of the fourregression equations, and all were coded as sets of categorical, or "dummy,11

variables. The set of greatest interest is the one describing the primary health

3. Respondents were asked to classify their plans by type, but were also asked to identify the plans byname. Plan names were used later to verify and, if necessary, correct the plan type given by therespondent.

4. The overall response rate to the health insurance supplement for the National Health Interview Surveywas about 92 percent

19

insurance plan that respondents had at the time of the survey. Three HMO categorieswere specified-* group or staff HMO, an IPA, and an HMO of unknown type. Theremaining insurance category contained all non-HMO plans, which would includeboth unmanaged and managed indemnity plans, such as those with utilization reviewprograms or with preferred provider organizations (PPOs). A number of additionalcategorical variables were included to control for factors other than insurance thatmight affect respondents1 use of medical services. They included variables fordelivery of a child during the year, race, age, sex, health status, education, income,region, and urbanization of the respondent's residence.5

The estimated coefficients were obtained using unweighted data, but theimplications of the regressions were calculated using weighted means to reflect thecharacteristics of the insured population.6 Table A-l lists the dependent andexplanatory variables used here, with their definitions and weighted sample means.Tables A-2 and A-3 show the estimated coefficients for the four regressions, alongwith their levels of significance.

Results

Estimated results were obtained separately for outpatient and inpatient services. Thedollar-weighted average of these separate effects was then used as an estimate of theoverall effects of HMOs on use of services.

Outpatient Visits. The first column in Table A-4 shows the implications derivedfrom the two regressions for use of outpatient visits. Compared with indemnityenrollees, enrollees in group/staff HMOs are 3 percent more likely to use someoutpatient services, although the number of services per user is not appreciablyhigher than it is for indemnity enrollees. Combining the findings for the probabilityof use and the extent of use for users, the results indicate that group/staff enrolleestypically use 3.2 percent more outpatient visits than similar indemnity enrollees.Although this effect is small, it is statistically significant-that is, the evidence is notconsistent with the hypothesis that group/staff enrollees use no more outpatient visitsthan similar indemnity enrollees.

5. A female respondent was identified as having given birth during the previous 12 months if her reportedhospital inpatient days excluding stays for delivery differed from the total inpatient days reported.

6. Because the NHIS uses a complex survey sampling scheme rather than simple random sampling, it isnecessary to use weighted data to produce representative estimates of the magnitude of events, butunweighted data produce reliable coefficient estimates.

20

TABLE A-l. VARIABLE DEFINITIONS AND WEIGHTED MEANS

Name of Variable

childbirth[excluded]

black[excluded]

chOOchl-6fe7-18fe!9-34fe35-54fe55-64ma7-18[excluded]ma35-54ma55-64

[excluded]hlth2hlth3hlth4hlthS

[excluded]educ2educ3educ4

[excluded]midncomehincomeunkncome

[excluded]region!region3region4

[excluded]msa2msa3msa4

grpstfipaunkhmo[excluded]

inpuseIn (inpdays)

outpuseIn (outpvsts)

Definition

1 if gave birth during year in hospital1 if did not give birth

1 if black1 if not black

1 if age is less than 11 if age is 1 through 61 if female age 7 through 181 if female age 19 through 341 if female age 35 through 541 if female age 55 through 641 if male age 7 through 181 if male age 19 through 341 if male age 35 through 541 if male age 55 through 64

1 if reported health is excellent1 if reported health is very good1 if reported health is good1 if reported health is fair1 if reported health is poor

1 if years of family head's education is under 121 if years of family head's education is 121 if years of family head's education is 13 through 161 if years of family head's education is 17 or more

1 if family income is under $35,0001 if family income is between $35,000 and $50,0001 if family income is $50,000 or more1 if family income is unreported

1 if residence is in Northeast1 if residence is in Midwest1 if residence is in South1 if residence is in West

1 if residence is an MSA-central city1 if residence is an MSA-not central city1 if residence is a nonfarm non-MSA1 if residence is a farm non-MSA

1 if primary health plan is a group or staff HMO1 if primary health plan is an IPA1 if primary health plan is an HMO, type unreported1 if primary health plan is indemnity

1 if respondent had any inpatient stays during yearnatural log of number of inpatient days for users

1 if respondent had any outpatient visits during yearnatural log of number of outpatient visits for users

Mean

0.0130.987

0.0960.904

0.0150.0930.0900.1340.1770.0550.0970.1250.1660.048

0.4540.3070.1900.0410.008

0.0440.3030.2620.390

0.3310.2240.3250.120

0.2340.2580.2780.230

0.2730.5480.1670.012

0.0590.1220.2200.599

0.0581.184

0.7930.974

SOURCE: Congressional Budget Office tabulations from the 1992 National Health Interview Survey.

NOTE: Except for the continuous dependent variables (inpdays and outpvsts), all are categorical, or"dummy," variables that have only two values—1 if the characteristic is present, and 0 if otherwise.Thus, the means show the proportion of people in the sample who have the specified characteristic.

21

TABLE A-2. REGRESSION ESTIMATES FOR OUTPATIENT VISITS

Logistic Regressionfor Probability of

Anv VisitsExplanatoryVariable8

interceptchildbirthblackchOOchl-6fe7-18fe!9-34fe35-54fe55-64ma7-18ma35-54ma55-64hlth2hlth3hlth4hlthSeduc2educ3educ4midncomehincomeunkncomeregion2regionsregion4msa2msa3msa4grpstfipaunkhmo

EstimatedCoefficient

-0.1892.114

-0.0962.3811.9550.7131.1370.8771.0430.7070.0560.4850.3680.5481.1561.8570.3950.6390.8420.0670.1950.021

-0.125-0.209-0.2410.019

-0.086-0.1830.1600.2430.208

P-valueb

0.00390.00010.01840.00010.00010.00010.00010.00010.00010.00010.14500.00010.00010.00010.00010.00010.00010.00010.00010.04010.00010.57940.00030.00010.00010.50420.01910.07060.00190.00010.0001

Least Squares Regressionfor Number of

VisitsEstimatedCoefficient

0.3991.319

-0.1540.6710.3360.0320.2570.1960.195

-0.0060.0590.0730.2340.5251.0211.7720.0820.1720.233

-0.014-0.001-0.1080.046

-0.0410.0290.016

-0.0040.0050.0010.0410.037

(Log Fonnl

P-valueb

0.00010.00010.00010.00010.00010.11110.00010.00010.00010.75810.00170.00430.00010.00010.00010.00010.00030.00010.00010.25170.93720.00010.00040.00130.02660.14060.75730.91290.93810.00360.0009

SOURCE: Congressional Budget Office regressions from the 1992 National Health Interview Survey.

a. See Table A-l for definitions.

b. The lower the reported P-value, the stronger (or more significant) the relationship is. All variables with aP-value less than 0.05 are generally considered to be significant.

22

TABLE A-3. REGRESSION ESTIMATES FOR INPATffiNT DAYS

Logistic Regressionfor Probability of

Anv DavsExplanatoryVariable"

interceptchildbirthblackchOOchl-6fe7-18fe!9-34fe35-54fe55-64ma7-18ma35-54ma55-64hlth2hltfaShlth4hlthSeduc2educSeduc4midncomehincomeunkncomeregion2regionSregion4msa2msa3rasa4grpstfipaunkhmo

EstimatedCoefficient

-3.9529.837

-0.0731.351

-0.127-0.3860.1870.3750.459

-0.4490.1790.6640.4051.0761.8542.7600.0090.1470.148

-0.122-0.113-0.2150.1490.028

-0.2140.1670.249

-0.023-0.443-0.0590.003

P-valueb

0.00010.00010.35320.00010.27880.00250.06200.00010.00010.00050.05850.00010.00010.00010.00010.00010.92760.14430.14720.05400.07670.00470.02230.66440.00290.00330.00050.91680.00020.43500.9590

Least Squares Regressionfor Number of

Pfty? (TL9e Fonn)EstimatedCoefficient

1.098-0.1400.0900.592

-0.1280.007

-0.0560.0230.1820.2640.1330.2820.0890.2030.4270.828

-0.115-0.031-0.082-0.024-0.0600.072

-0.033-0.065-0.1460.0330.073

-0.1890.010

-0.015-0.021

P-valueb

0.00010.00870.11240.00010.18810.94790.46980.74640.03560.01300.08760.00130.04340.00010.00010.00010.10350.67370.26430.59210.19060.18320.47480.16130.00360.41800.15370.23720.89720.76750.6085

SOURCE: Congressional Budget Office regressions from the 1992 National Health Interview Survey.

a. See Table A-l for definitions.

b. The lower the reported P-value, the stronger (or more significant) the relationship is. All variables with aP-value less than 0.05 are generally considered to be significant.

23

TABLE A-4. ESTIMATED CHANGE IN USE OF SERVICES FOR HMOsCOMPARED WITH TYPICAL INDEMNITY PLANS

Percenta&e Chanee in UseOutpatient

VisitsInpatient

DaysMedicalServices8

HMOsGroup/StaffIPAUnreported

HMO Average15


HMO Average1*

Probability of Any Use

3.04.53.9

3.9

Extent of Use Among Users

0.14.13.7

2.6

Total Use of Services

-34.8-5.50.3

-16.3

1.0-1.5-2.1

-0.6

-20.1-1.61.7

-8.4

0.70.602

0.7


HMO Averageb

3.28.77.7

6.7

-34.2-6.9-1.8

-17.0

-19.6-0.81.9

-7.8

SOURCE: Congressional Budget Office estimates from the 1992 National Health Interview Survey.

NOTES: Typical indemnity plan refers to a fee-for-service plan with some elements of managed care.HMO = health maintenance organization; IPA = independent practice association.

a. Calculation of effects on use of medical services weights outpatient visits by 0.39 and inpatient days by0.61, to reflect the mix of spending on outpatient and inpatient services.

b. Calculation of the HMO average uses weights of 0.37 for group/staff HMOs, 0.63 for IPAs, and 0 forHMOs of unreported type.

24

The increase in use of outpatient visits for other HMO enrollees, comparedwith indemnity enrollees, is 8 percent to 9 percent-and is thus larger than it is forgroup/staff enrollees. Further, the higher use among other HMO enrollees results asmuch from more visits per user as from a higher probability of any use, in contrastto the results for group/staff HMOs, in which most of the effect was due to a higherprobability of any use during the year. The probability of any use is primarily theresult of patients1 preferences and would be expected to be higher among HMOenrollees because of the lower cost-sharing requirements typical in HMOs. Theextent of use once a patient has initiated contact with the medical system, however,is a function of both the patient's preferences and the provider's responses. Thedifference between group/staff HMOs and other HMOs concerning the extent of useamong users is indicative of the more effective control that group/staff HMOsapparently have over providers' treatment patterns.

Hospital Inpatient Days. The second column in Table A-4 shows the implicationsderived from the two regressions for use of inpatient hospital days. On average, theprobability of an inpatient stay is about 35 percent lower for enrollees in group/staffHMOs than for similar people in indemnity plans. Among those who are admittedto the hospital, the number of days used is slightly (and not significantly) higher forgroup/staff HMO enrollees than for similar people in indemnity plans-perhapsbecause the much lower admission rate means that those who are admitted are sicker.Total use of hospital inpatient days is about 34 percent lower for group/staff HMOenrollees compared with similar people in indemnity plans.

The estimates indicate that IPA enrollees use about 7 percent fewer hospitalinpatient days than similar people in indemnity plans, with most of this effectattributable to fewer hospital admissions. However, these effects are not statisticallysignificant. In other words, these results are consistent with the hypothesis that IP As,on average, do not reduce the use of inpatient days compared with indemnity plans.

Among HMO enrollees who do not know whether their plan is a group/staffmodel or an IPA, the estimated effects on use of inpatient days are even smaller.Compared with indemnity enrollees, this group uses about 2 percent fewer inpatientdays. Here, too, the difference is not statistically significant.

Overall Use of Medical Services. The results discussed above are combined in thethird column in Table A-4 to obtain an estimate of the effects of HMOs on theresource cost of medical services overall.7 For this estimate, it is assumed that the

7. Nominal costs would include the effects of prices paid per service as well, but that is beyond the scopeof this analysis. The NHIS does not include information on health care expenditures.

25

resource costs of all outpatient care are proportional to the number of outpatient visitsmade, and that the costs of all inpatient care are proportional to the number ofinpatient days used. For the population under age 65 in 1987, about 39 percent ofspending on insured services was for outpatient care, and 61 percent was for inpatientcare.8 These values are used to weight the results discussed above to estimate theeffects of HMOs on the overall use of medical services. The estimates given herefor overall use of medical services would probably be somewhat different if laterexpenditure data were available.

For each HMO group, use of outpatient services is somewhat higher than itwould be for similar enrollees in indemnity plans, whereas use of inpatient servicesis lower. For group/staff HMOs, the combined effect on use of medical servicesoverall is a substantial reduction of nearly 20 percent when compared with indemnityplans, because the reduction in inpatient use is large and the increase in outpatientuse is small. For other HMO groups, small reductions in inpatient use are largelyoffset by increases in outpatient use of services, with little or no change in use ofmedical services overall. The average effect on overall use of medical services isa reduction of 7.8 percent

Comparison with Other Research

Only one of the previously published studies that compares the effects of group/staffHMOs and IP As with indemnity plans is based on nationally representative data.The other studies are based on selected plans, physicians, and patients that are notnationally representative.9 Because it is inappropriate to conclude that theperformance of HMOs in studies of selected plans is typical of all HMOs, thiscomparison focuses only on the nationally representative study by Lewin-VHI.

In addition to the different survey year examined (1989 instead of 1992), themethods used for the Lewin-VHI study differed from those used here in severalways. One important difference is that this analysis included an indicator forchildbirth in the regressions, whereas the Lewin-VHI study did not. Otherdifferences in specification that may have contributed to different results are that the

8. Percentages are based on tabulations from the 1987 National Medical Expenditure Survey, the latestexpenditure data available. Both outpatient and inpatient expenditures include facility, physician, andother professional costs.

9. For example, see Sheldon Greenfield and others, "Variations in Resource Utilization Among MedicalSpecialties and Systems of Care,11 Journal of the American Medical Association, vol. 267, no. 12(March 25,1992); and Randall Brown and Jerrold Hill, MDoes Model Type Play a Role in the Extentof HMO Effectiveness in Controlling the Utilization of Services?*1 (Princeton, N.J.: MathematicaPolicy Research, Inc., May 10,1993).

26

Lewin-VHI study used only a single ordinary least squares regression to estimatetotal use of outpatient visits, instead of the two-part estimation procedure used here;further, it included uninsured and publicly insured people in the sample along withprivately insured people, combined those in HMOs of unspecified type with those inIP As, and estimated separate equations for children and adults.

For both group/staff HMOs and IP As, the reduction in total use of servicesestimated here is about twice as large as the estimates reported by Lewin-VHI (seeresults for total use of services in Table A-4, in comparison with Table A-5).Group/staff HMOs reduce use of medical services by nearly 20 percent, instead ofthe 9 percent reduction implied by the Lewin-VHI results. IP As reduce use by about0.8 percent, instead of 0.3 percent. Averaged over both types, these estimatesindicate that HMOs reduced use of services by nearly 8 percent in 1992, whencompared with indemnity plans. By contrast, the Lewin-VHI results imply areduction of 3.9 percent overall using the 1989 mix of HMOs, or a reduction of 3.6percent using the 1992 mix.

When the explanatory variable for childbirth is excluded from the regressionsin this analysis, the estimated reduction in use for HMOs overall is less than half theestimated drop when the childbirth variable is included (see Table A-6). Theestimated reduction in use of inpatient services is only about half as large, while theincrease in use of outpatient visits is slightly larger. Thus, excluding the childbirthvariable produces estimates much closer to the results reported by Lewin-VHI. Thisprobably indicates that the absence of a variable for childbirth in the data used byLewin-VHI produced an estimate of HMO effects on reducing use of inpatient daysthat was too small, so the estimated effect on use of medical services overall was alsotoo small.

27

TABLE A-5. PREVIOUS ESTIMATES OF CHANGE IN USE OF SERVICESFOR HMOs COMPARED WITH TYPICAL INDEMNITY PLANS

Percentage Change i

HMOsGroup/StaffIP A or Unreported

HMO AverageAs Reported5

Using 1992 Weights0

OutpatientVisits

Total Use of Services

6.69.9

8.48.7

InpatientDays

-19.0-6.9

-11.7-11.4

D USC

MedicalServices8

-9.1-0.3

-3.9-3.6

SOURCE: Congressional Budget Office, derived from Table A-4 in "The Financial Impact of the HealthSecurity Act" (Fairfax, Va.: Lewin-Vffl, Inc., December 9,1993).


a. Results for medical services were not reported in the study, but were calculated from reported results foroutpatient visits and inpatient days using 0.39 and 0.61 as the respective weights.

b. Calculation of the HMO average as reported uses (1989) weights of 0.41 for group/staff HMOs and 0.59for other HMOs.

c. Calculation of the HMO average for 1992 uses weights of 0.37 for group/staff HMOs and 0.63 for otherHMOs.

28

TABLE A-6. COMPARISON UNDER ALTERNATIVE SPECIFICATIONSOF ESTIMATED CHANGE IN USE OF SERVICES FOR HMOsCOMPARED WITH TYPICAL INDEMNITY PLANS

Percentage Chanee in UseOutpatient Inpatient Medical

Visits Days Services8

CBO Model with Childbirth Variable (Total Use of Services)

HMOsGroup/Staff 3.2 -34.2 -19.6IPA 8.7 -6.9 -0.8Unreported 7.7 -1.8 1.9

HMO Average6 6.7 -17.0 -7.8

CBO Model Without Childbirth Variable (Total Use of Services)


HMOAverageb

3.4928.1

7.0

-23.5-1.91.4

-9.9

-13.02.44.0

-3.3

SOURCE: Congressional Budget Office estimates from the 1992 National Health Interview Survey.


a. Calculation of effects on use of medical services weights outpatient visits by 0.39 and inpatient days by0.61, to reflect the mix of spending on outpatient and inpatient services.

b. Calculation of the HMO average uses weights of 0.37 for group/staff HMOs, 0.63 for IPAs, and 0 forHMOs of unreported type.

29

Documents

CBO MEMORANDUM